seemingly converged solution changes after restart

achat · September 3, 2013, 10:03

Hello:

I was wondering if anyone encountered the following, rather wierd, problem and if anyone could offer help/suggestions. I am using FLUENT.

This is a low pressure CVD problem of fiber coating with surface reactions. The solution is steady state. The 3D equation being solved are momentum, energy (with radiation) and a few species concentrations. For a few cases what I find is that when I restart a seemingly converged solution without changing anything, the steady state solution converges to a different state. Before the restart, the seemingly converged showed all traits of convergence: the weight gains had plateaued out, the residuals were small (<1e-4) and plateaued and so on. Also if I allow the seemingly converged solution to run much longer without interruption it would remain in the same (seemingly) converged state.

Clearly interrupting and restarting is doing something; as if waking the solution out of a metastable state. Has anyone encountered such a problem? What I also find is that for these cases when I restart, the residuals spike quite a bit although I am not changing anything in the restart operation. This too is surprising.

Thanks

Athan

Totalsim · September 3, 2013, 10:57

Is it restarting from scratch or does it 'blip'?

It could be something as simple as the precision it is written out in. If you are solving in double and writing out in single precision that might be your problem.

Good luck.

Rockfeaver · September 3, 2018, 03:55

Hi, everybody!

I have the same problem. I would say it could be due to change of Fluent version or usage of different number of nodes (if you calculate in cluster), as my colleagues said. However, it's seems not to be the case. I used the same fluent version and same number of nodes all the time.

So, if some of you guys found a solution, please give us a feedback.

Thank you in advance!

September 3, 2013, 10:03	seemingly converged solution changes after restart	#1
achat New Member athan Join Date: Aug 2013 Posts: 2 Rep Power: 0	Hello: I was wondering if anyone encountered the following, rather wierd, problem and if anyone could offer help/suggestions. I am using FLUENT. This is a low pressure CVD problem of fiber coating with surface reactions. The solution is steady state. The 3D equation being solved are momentum, energy (with radiation) and a few species concentrations. For a few cases what I find is that when I restart a seemingly converged solution without changing anything, the steady state solution converges to a different state. Before the restart, the seemingly converged showed all traits of convergence: the weight gains had plateaued out, the residuals were small (<1e-4) and plateaued and so on. Also if I allow the seemingly converged solution to run much longer without interruption it would remain in the same (seemingly) converged state. Clearly interrupting and restarting is doing something; as if waking the solution out of a metastable state. Has anyone encountered such a problem? What I also find is that for these cases when I restart, the residuals spike quite a bit although I am not changing anything in the restart operation. This too is surprising. Thanks Athan

September 3, 2013, 10:57		#2
Totalsim Member Jon Join Date: Mar 2013 Posts: 47 Rep Power: 13	Is it restarting from scratch or does it 'blip'? It could be something as simple as the precision it is written out in. If you are solving in double and writing out in single precision that might be your problem. Good luck. __________________ TotalSim CFD Engineer www.totalsimulation.co.uk

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
grid dependancy	gueynard a.	Main CFD Forum	19	June 27, 2014 21:22
Can this solution be regarded as converged?	Chander	CFX	3	April 3, 2011 06:42
loading new mesh in converged solution	Claudia	FLUENT	1	November 4, 2005 05:23
Has the pressure correction method problems?	X. Ye	Main CFD Forum	17	September 1, 1999 16:14
Wall functions	Abhijit Tilak	Main CFD Forum	6	February 5, 1999 01:16

September 3, 2018, 03:55		#3
Rockfeaver New Member Yury Voloshchuk Join Date: Aug 2017 Posts: 2 Rep Power: 0	Hi, everybody! I have the same problem. I would say it could be due to change of Fluent version or usage of different number of nodes (if you calculate in cluster), as my colleagues said. However, it's seems not to be the case. I used the same fluent version and same number of nodes all the time. So, if some of you guys found a solution, please give us a feedback. Thank you in advance!